Deep and shallow work
Cal Newport draws a line between cognitively demanding work performed without distraction, which he calls deep work, and everything else. Designing the structure of a report is the first kind; emailing it around is the second.
The point of the distinction is not to look down on shallow work. It is unavoidable and it will occupy a large share of any job. The problem is that it tends to be sprinkled evenly across the day, leaving no continuous space for the other kind.
Attention residue
When you switch tasks, part of your attention stays with the previous one. Check email for five minutes and the cost is not five minutes - it is five minutes plus however long it takes to be fully back.
That makes 'just checking' expensive. A day spent glancing at notifications every half hour can contain almost no deep work at all, and frequent interruption means you never descend far enough to do more than skim the surface of a problem.
A procedure that works
Do not try to convert the whole day. Two protected hours is a large change in most jobs.
Book the time first
Work that happens 'when there is time' does not happen. Put a recurring block on the calendar and defend it against other invitations. Keeping it at the same hour each day helps, because the routine itself lowers the cost of getting started.
Cut the interruption paths physically
Willpower is a poor filter. Log out rather than muting, put the phone in another room, close the tabs. Adding a few seconds of friction removes most impulses.
Interruptions from colleagues are largely solved by announcing the block in advance. 'Slow to reply between 10 and 12' is usually accepted without comment.
Write down self-interruptions
You will interrupt yourself more often than anyone else does. When something occurs to you - a thing to look up, an errand - write it down and keep going. Reading the list afterwards makes it obvious how little of it was urgent.
Give it an end
Depth is not an unlimited resource. Untrained, an hour is demanding; even with practice, three or four hours a day is the realistic ceiling. Setting an explicit end with a timer raises the density inside it.
Batch the shallow work
A protected block only survives if the shallow work has somewhere else to go. Handle mail and messages at two or three fixed points, and cluster meetings in the afternoon. Deep work in the front half of the day and shallow work in the back half is a layout that holds up well.
Counterintuitively, the shallow work gets finished faster this way, because batching removes most of the switching cost.